Re: LC comment for SpecGL : 'what does "MUST define scope" mean?'

On Thu, 24 Apr 2003, Dan Connolly wrote:

> Yes, and since it's not testable, it seems counterproductive to
> phrase it using rfc2119:MUST.

A significant percentage of existing MUSTs (across all RFCs, etc.) are
not testable, but still essential or, at least, very useful. Note that
"MUST" definition in RFC 2119 says nothing about testability.

I do not have any accurate stats, but my experience shows that at
least 20% of MUSTs are not testable at all and at least 99% are not
fully testable (i.e., no finite algorithm can identify compliant
implementations correctly). This is the reality: RFCs are written in
imprecise language and implementations are not formal models either.

> > RFC2119 scope is not limited to specs about software agents.
>
> OK, so never mind software; yes, RFC2119 constraints can be applied
> to any sort of agent: software, a person, a group, or some
> combination. But a spec is none of these.

RFC2119 scope is not limited to specs about agents. For example, specs
about interfaces (such as XML) can and do use RFC2119. A spec is an
interface itself. Alternatively, I can say that our "agents" are spec
readers and spec authors.

> In case it's not clear, I'm not satisfied by this response to my
> last call comment.

IMHO, your objections go against current [best] practices/norms. This
is a murky area where no precise answers/definitions are possible
because we are dealing with generally imprecise/informal objects
(specs). I hope my response clarifies my position (I may have proposed
adding that MUST, not sure). I do thank you for submitting the
comment. I will have to leave the issue for the working group to
resolve. I am not a WG member, and I do not even know whether your
satisfaction is guaranteed by the W3C process.

Thank you,

Alex.

Received on Thursday, 24 April 2003 10:52:13 UTC